fix(windows): pin the search_code PowerShell pipe to UTF-8

PowerShell 5.1 encodes stdout for a native-process pipe in the console
OEM codepage, so raw search content containing characters the inherited
CP cannot carry (Cyrillic under CP437/850, ...) reached
collect_grep_matches as '?' — and whether it degraded depended entirely
on which console the server happened to inherit. That surfaced as the
intermittent test_mcp raw-Русский mojibake on the windows CI leg and
means real Windows users in a default console get '?' for all
non-ASCII search_code raw content.

Every generated command now pins [Console]::OutputEncoding to UTF-8, so
the pipe is codepage-independent by construction. The read side needs
no pin: Select-String decodes BOM-less UTF-8 via .NET StreamReader
defaults. A Windows-side builder test asserts all five command variants
carry the prelude.

Signed-off-by: Martin Vogel <martin.vogel.tech@gmail.com>

src/mcp/mcp.c

@@ -8848,6 +8848,16 @@ bool cbm_search_code_file_pattern_can_prefilter(const char *file_pattern) {
 /* Build the grep/search command string based on scoped vs recursive mode.
  * On Windows, uses PowerShell Select-String with tab-delimited output.
  * On POSIX, uses grep with colon-delimited output. */
+/* Windows PowerShell 5.1 encodes stdout for a native-process pipe in the
+ * console OEM codepage, so any character the inherited CP cannot carry
+ * (Cyrillic under CP437/850, etc.) degrades to '?' before it ever reaches
+ * collect_grep_matches — and WHETHER it degrades depends on the console the
+ * server happened to inherit. Pin the pipe to UTF-8 inside every generated
+ * command so raw search content is codepage-independent. (The read side is
+ * already safe: Select-String decodes BOM-less UTF-8 via .NET StreamReader
+ * defaults.) */
+#define CBM_PS_UTF8_PRELUDE "[Console]::OutputEncoding=[System.Text.Encoding]::UTF8; "
+
 void cbm_search_code_build_grep_cmd(char *cmd, size_t cmd_sz, bool use_regex, bool scoped,
                                     const char *file_pattern, const char *tmpfile,
                                     const char *filelist, const char *root_path) {
@@ -8858,7 +8868,8 @@ void cbm_search_code_build_grep_cmd(char *cmd, size_t cmd_sz, bool use_regex, bo
             if (cbm_search_code_file_pattern_can_prefilter(file_pattern)) {
                 snprintf(
                     cmd, cmd_sz,
-                    "powershell -Command \"$pat = Get-Content -Encoding UTF8 -LiteralPath '%s'; "
+                    "powershell -Command \"" CBM_PS_UTF8_PRELUDE
+                    "$pat = Get-Content -Encoding UTF8 -LiteralPath '%s'; "
                     "Get-Content -Encoding UTF8 -LiteralPath '%s'"
                     " | Where-Object { $_ -like '%s' }"
                     " | ForEach-Object { Select-String -LiteralPath $_ -Pattern $pat%s "
@@ -8869,7 +8880,8 @@ void cbm_search_code_build_grep_cmd(char *cmd, size_t cmd_sz, bool use_regex, bo
             } else {
                 snprintf(
                     cmd, cmd_sz,
-                    "powershell -Command \"$pat = Get-Content -Encoding UTF8 -LiteralPath '%s'; "
+                    "powershell -Command \"" CBM_PS_UTF8_PRELUDE
+                    "$pat = Get-Content -Encoding UTF8 -LiteralPath '%s'; "
                     "Get-Content -Encoding UTF8 -LiteralPath '%s' | ForEach-Object { Select-String "
                     "-LiteralPath $_ -Pattern $pat%s "
                     "-ErrorAction SilentlyContinue }"
@@ -8880,7 +8892,8 @@ void cbm_search_code_build_grep_cmd(char *cmd, size_t cmd_sz, bool use_regex, bo
         } else {
             snprintf(
                 cmd, cmd_sz,
-                "powershell -Command \"$pat = Get-Content -Encoding UTF8 -LiteralPath '%s'; "
+                "powershell -Command \"" CBM_PS_UTF8_PRELUDE
+                "$pat = Get-Content -Encoding UTF8 -LiteralPath '%s'; "
                 "Get-Content -Encoding UTF8 -LiteralPath '%s' | ForEach-Object { Select-String "
                 "-LiteralPath $_ -Pattern $pat%s "
                 "-ErrorAction SilentlyContinue }"
@@ -8891,7 +8904,8 @@ void cbm_search_code_build_grep_cmd(char *cmd, size_t cmd_sz, bool use_regex, bo
         if (file_pattern) {
             snprintf(
                 cmd, cmd_sz,
-                "powershell -Command \"Get-ChildItem -Recurse -Path '%s\\*' -Include '%s' -File "
+                "powershell -Command \"" CBM_PS_UTF8_PRELUDE
+                "Get-ChildItem -Recurse -Path '%s\\*' -Include '%s' -File "
                 "-ErrorAction SilentlyContinue"
                 " | Select-String -Pattern (Get-Content -Encoding UTF8 -LiteralPath '%s')%s "
                 "-ErrorAction SilentlyContinue"
@@ -8900,7 +8914,8 @@ void cbm_search_code_build_grep_cmd(char *cmd, size_t cmd_sz, bool use_regex, bo
         } else {
             snprintf(
                 cmd, cmd_sz,
-                "powershell -Command \"Get-ChildItem -Recurse -Path '%s\\*' -File -ErrorAction "
+                "powershell -Command \"" CBM_PS_UTF8_PRELUDE
+                "Get-ChildItem -Recurse -Path '%s\\*' -File -ErrorAction "
                 "SilentlyContinue"
                 " | Select-String -Pattern (Get-Content -Encoding UTF8 -LiteralPath '%s')%s "
                 "-ErrorAction SilentlyContinue"

tests/test_mcp.c

@@ -2120,7 +2120,7 @@ TEST(tool_search_graph_includes_node_properties) {
     ASSERT_NOT_NULL(strstr(inner, "func HandleRequest")); /* its value */
     ASSERT_NOT_NULL(strstr(inner, "\"base_classes\""));
     ASSERT_NOT_NULL(strstr(inner, "[\"HandlerBase\",\"Audited\"]"));
-    ASSERT_NULL(strstr(inner, "is_exported"));            /* blob never spills */
+    ASSERT_NULL(strstr(inner, "is_exported")); /* blob never spills */
     free(inner);
     free(resp);
 
@@ -2680,15 +2680,13 @@ TEST(tool_check_index_coverage_accepts_truncated_ignored_catalog_for_fresh_path_
         ((int64_t)source_stat.st_mtimespec.tv_sec * (int64_t)CBM_NSEC_PER_SEC) +
         (int64_t)source_stat.st_mtimespec.tv_nsec;
 #elif defined(_WIN32)
-    int64_t source_mtime_ns =
-        (int64_t)source_stat.st_mtime * (int64_t)CBM_NSEC_PER_SEC;
+    int64_t source_mtime_ns = (int64_t)source_stat.st_mtime * (int64_t)CBM_NSEC_PER_SEC;
 #else
-    int64_t source_mtime_ns =
-        ((int64_t)source_stat.st_mtim.tv_sec * (int64_t)CBM_NSEC_PER_SEC) +
-        (int64_t)source_stat.st_mtim.tv_nsec;
+    int64_t source_mtime_ns = ((int64_t)source_stat.st_mtim.tv_sec * (int64_t)CBM_NSEC_PER_SEC) +
+                              (int64_t)source_stat.st_mtim.tv_nsec;
 #endif
-    ASSERT_EQ(cbm_store_upsert_file_hash(store, "test-project", "main.go", "",
-                                         source_mtime_ns, source_stat.st_size),
+    ASSERT_EQ(cbm_store_upsert_file_hash(store, "test-project", "main.go", "", source_mtime_ns,
+                                         source_stat.st_size),
               CBM_STORE_OK);
     cbm_project_t project = {0};
     ASSERT_EQ(cbm_store_get_project(store, "test-project", &project), CBM_STORE_OK);
@@ -2702,8 +2700,7 @@ TEST(tool_check_index_coverage_accepts_truncated_ignored_catalog_for_fresh_path_
         .coverage_version = 1,
         .hash_records_complete = true,
     };
-    ASSERT_EQ(cbm_store_coverage_replace_ex(store, "test-project", NULL, 0, &meta),
-              CBM_STORE_OK);
+    ASSERT_EQ(cbm_store_coverage_replace_ex(store, "test-project", NULL, 0, &meta), CBM_STORE_OK);
     cbm_project_free_fields(&project);
 
     char *response = cbm_mcp_handle_tool(
@@ -4351,12 +4348,12 @@ TEST(search_code_full_preserves_utf8_source) {
                           .end_line = 2};
     ASSERT_GT(cbm_store_upsert_node(st, &section), 0);
 
-    char *resp = cbm_mcp_server_handle(
-        srv, "{\"jsonrpc\":\"2.0\",\"id\":97,\"method\":\"tools/call\","
-             "\"params\":{\"name\":\"search_code\",\"arguments\":{"
-             "\"project\":\"utf8-search\",\"pattern\":\"accounting-design\","
-             "\"file_pattern\":\"*.md\",\"path_filter\":\"^design/\","
-             "\"mode\":\"full\",\"format\":\"json\",\"limit\":5}}}");
+    char *resp =
+        cbm_mcp_server_handle(srv, "{\"jsonrpc\":\"2.0\",\"id\":97,\"method\":\"tools/call\","
+                                   "\"params\":{\"name\":\"search_code\",\"arguments\":{"
+                                   "\"project\":\"utf8-search\",\"pattern\":\"accounting-design\","
+                                   "\"file_pattern\":\"*.md\",\"path_filter\":\"^design/\","
+                                   "\"mode\":\"full\",\"format\":\"json\",\"limit\":5}}}");
     ASSERT_NOT_NULL(resp);
     char *inner = extract_text_content(resp);
     ASSERT_NOT_NULL(inner);
@@ -4408,11 +4405,11 @@ TEST(search_code_raw_match_preserves_utf8_content) {
                        .end_line = 1};
     ASSERT_GT(cbm_store_upsert_node(st, &node), 0);
 
-    char *resp = cbm_mcp_server_handle(
-        srv, "{\"jsonrpc\":\"2.0\",\"id\":98,\"method\":\"tools/call\","
-             "\"params\":{\"name\":\"search_code\",\"arguments\":{"
-             "\"project\":\"test-project\",\"pattern\":\"raw-\","
-             "\"file_pattern\":\"*.md\",\"format\":\"json\",\"limit\":5}}}");
+    char *resp =
+        cbm_mcp_server_handle(srv, "{\"jsonrpc\":\"2.0\",\"id\":98,\"method\":\"tools/call\","
+                                   "\"params\":{\"name\":\"search_code\",\"arguments\":{"
+                                   "\"project\":\"test-project\",\"pattern\":\"raw-\","
+                                   "\"file_pattern\":\"*.md\",\"format\":\"json\",\"limit\":5}}}");
     ASSERT_NOT_NULL(resp);
     char *inner = extract_text_content(resp);
     ASSERT_NOT_NULL(inner);
@@ -4459,11 +4456,11 @@ TEST(search_code_context_preserves_utf8_context) {
                        .end_line = 3};
     ASSERT_GT(cbm_store_upsert_node(st, &node), 0);
 
-    char *resp = cbm_mcp_server_handle(
-        srv, "{\"jsonrpc\":\"2.0\",\"id\":99,\"method\":\"tools/call\","
-             "\"params\":{\"name\":\"search_code\",\"arguments\":{"
-             "\"project\":\"test-project\",\"pattern\":\"context-needle\","
-             "\"format\":\"json\",\"context\":1,\"limit\":5}}}");
+    char *resp =
+        cbm_mcp_server_handle(srv, "{\"jsonrpc\":\"2.0\",\"id\":99,\"method\":\"tools/call\","
+                                   "\"params\":{\"name\":\"search_code\",\"arguments\":{"
+                                   "\"project\":\"test-project\",\"pattern\":\"context-needle\","
+                                   "\"format\":\"json\",\"context\":1,\"limit\":5}}}");
     ASSERT_NOT_NULL(resp);
     char *inner = extract_text_content(resp);
     ASSERT_NOT_NULL(inner);
@@ -4507,8 +4504,8 @@ TEST(search_code_invalid_utf8_still_returns_valid_json) {
     char invalid_path[512];
     snprintf(invalid_path, sizeof(invalid_path), "%s/project/invalid.md", tmp);
     static const unsigned char invalid_source[] = {
-        'i', 'n', 'v', 'a', 'l', 'i', 'd', '-', 'n', 'e', 'e', 'd', 'l', 'e', '\n',
-        'c', 'o', 'n', 't', 'e', 'x', 't', ' ', 0xFF, '\n',
+        'i', 'n',  'v', 'a', 'l', 'i', 'd', '-', 'n', 'e', 'e',  'd',  'l',
+        'e', '\n', 'c', 'o', 'n', 't', 'e', 'x', 't', ' ', 0xFF, '\n',
     };
     FILE *fp = cbm_fopen(invalid_path, "wb");
     ASSERT_NOT_NULL(fp);
@@ -4526,11 +4523,11 @@ TEST(search_code_invalid_utf8_still_returns_valid_json) {
                        .end_line = 2};
     ASSERT_GT(cbm_store_upsert_node(st, &node), 0);
 
-    char *resp = cbm_mcp_server_handle(
-        srv, "{\"jsonrpc\":\"2.0\",\"id\":100,\"method\":\"tools/call\","
-             "\"params\":{\"name\":\"search_code\",\"arguments\":{"
-             "\"project\":\"test-project\",\"pattern\":\"invalid-needle\","
-             "\"mode\":\"full\",\"format\":\"json\",\"limit\":5}}}");
+    char *resp =
+        cbm_mcp_server_handle(srv, "{\"jsonrpc\":\"2.0\",\"id\":100,\"method\":\"tools/call\","
+                                   "\"params\":{\"name\":\"search_code\",\"arguments\":{"
+                                   "\"project\":\"test-project\",\"pattern\":\"invalid-needle\","
+                                   "\"mode\":\"full\",\"format\":\"json\",\"limit\":5}}}");
     ASSERT_NOT_NULL(resp);
     char *inner = extract_text_content(resp);
     ASSERT_NOT_NULL(inner);
@@ -4921,6 +4918,35 @@ TEST(search_code_windows_prefilter_precedes_content_scan) {
 #endif
 }
 
+/* Windows raw scans must pin the PowerShell pipe to UTF-8: PS 5.1 otherwise
+ * emits stdout in the console OEM codepage and non-ASCII content degrades to
+ * '?' depending on the inherited console CP (seen as test_mcp raw-Русский
+ * mojibake on CI). Every builder variant must carry the prelude. */
+TEST(search_code_windows_scan_pins_utf8_output) {
+#ifdef _WIN32
+    static const char prelude[] =
+        "powershell -Command \"[Console]::OutputEncoding=[System.Text.Encoding]::UTF8; ";
+    char command[CBM_SZ_4K];
+    struct {
+        bool scoped;
+        const char *file_pattern;
+    } cases[] = {{true, "*.go"},     /* scoped + prefilterable pattern */
+                 {true, "*x*z*.go"}, /* scoped + non-prefilterable pattern */
+                 {true, NULL},       /* scoped, no pattern */
+                 {false, "*.go"},    /* unscoped + pattern */
+                 {false, NULL}};     /* unscoped, no pattern */
+    for (size_t i = 0; i < sizeof(cases) / sizeof(cases[0]); i++) {
+        cbm_search_code_build_grep_cmd(command, sizeof(command), false, cases[i].scoped,
+                                       cases[i].file_pattern, "C:/tmp/pattern", "C:/tmp/filelist",
+                                       "C:/tmp/root");
+        ASSERT_TRUE(strncmp(command, prelude, sizeof(prelude) - 1) == 0);
+    }
+    PASS();
+#else
+    SKIP_PLATFORM("PowerShell scan encoding applies on Windows");
+#endif
+}
+
 /* issue #283: search_code with regex=true and a syntactically invalid pattern
  * must return an explicit error, not an empty result indistinguishable from a
  * legitimate no-match. */
@@ -8769,10 +8795,10 @@ TEST(tool_resolve_store_by_internal_name_issue704) {
     ASSERT_NULL(strstr(list, "ghost704"));     /* 0-byte ghost filtered (RED before) */
     free(list);
 
-    char *page = cbm_mcp_server_handle(
-        srv, "{\"jsonrpc\":\"2.0\",\"id\":11,\"method\":\"tools/call\","
-             "\"params\":{\"name\":\"list_projects\","
-             "\"arguments\":{\"offset\":0,\"limit\":1}}}");
+    char *page =
+        cbm_mcp_server_handle(srv, "{\"jsonrpc\":\"2.0\",\"id\":11,\"method\":\"tools/call\","
+                                   "\"params\":{\"name\":\"list_projects\","
+                                   "\"arguments\":{\"offset\":0,\"limit\":1}}}");
     ASSERT_NOT_NULL(page);
     ASSERT_NOT_NULL(strstr(page, "\\\"total\\\":3"));
     ASSERT_NOT_NULL(strstr(page, "\\\"limit\\\":1"));
@@ -11223,6 +11249,7 @@ SUITE(mcp) {
     RUN_TEST(search_code_path_filter_matches_nothing);
     RUN_TEST(search_code_file_pattern_prefilter_boundaries);
     RUN_TEST(search_code_windows_prefilter_precedes_content_scan);
+    RUN_TEST(search_code_windows_scan_pins_utf8_output);
     RUN_TEST(search_code_invalid_regex_errors_issue283);
     RUN_TEST(search_code_literal_pipe_warns_issue282);
     RUN_TEST(search_code_reports_phase_timings_only_in_debug_mode);